Abstract: After analyzing Internet communication delay components and features，the Elman neural network is adopted to predict the Internet communication delay.The prediction is simulated using Matlab.The results show that the Elman neural network can provide a promising way to predict the Internet timedelay.In order to improve the dynamical characteristics and system ability of the neural network，a modified neural dynamic recurrent network based on input，output and hide layers is proposed.Simulation results show that the modified Elman neural network has better dynamic performance.

摘要： 分析了网络传输时延的组成和特点，提出了利用Elman神经网络预测网络传输时延，运用Matlab软件对其预测进行仿真，结果证明Elman神经网络能很好地预测网络时延，为了进一步提高神经网络的逼近能力和动态特性，提出了一种改进的基于输入层、隐藏层、输出层神经元的动态递归神经网络。实验证明，改进的Elman神经网络比原来的网络具有更好的动态性能。